364
The International Neuroblastoma Pathology
Classification (the Shimada System)
Hiroyuki Shimada, M.D., Ph.D.1
Inge M. Ambros, M.D.2
Louis P. Dehner, M.D.3
Jun-ichi Hata, M.D.4
Vijay V. Joshi, M.D.5
Borghild Roald, M.D.6
Daniel O. Stram, Ph.D.7
Robert B. Gerbing, M.A.7
John N. Lukens, M.D.8
Katherine K. Matthay, M.D.9
Robert P. Castleberry, M.D.10
1
Department of Pathology and Laboratory Medicine,
Childrens Hospital Los Angeles, Los Angeles, California.
2
Children’s Cancer Research Institute, St. Anna
Kinderspital, Vienna, Austria.
3
Department of Pathology, Barnes Hospital, St.
Louis, Missouri.
4
Department of Pathology, Keio University School
of Medicine, Tokyo, Japan.
5
Department of Pathology and Laboratory Medi-
cine, Hartford Hospital, Hartford, Connecticut.
6
Department of Pathology, Ulleval University Hos-
pital, Oslo, Norway.
7
The Childrens Cancer Group, Arcadia, California.
8
Pediatric Hematology/Oncology, Vanderbilt Uni-
versity Medical Center, Nashville, Tennessee.
9
Pediatric Hematology/Oncology, University of
California San Francisco, School of Medicine, San
Francisco, California.
10
Department of Pediatrics, University of Alabama
at Birmingham, Birmingham, Alabama.
Presented in part at the Annual Meeting of the Amer-
ican Society of Clinical Oncology, Denver, Colorado,
May 17–20, 1997, and at the Annual Meeting of the
Society for Pediatric Pathology, Boston, Massachu-
setts, February 28–March 1, 1998.
Supported in part by contributions from the William
G. Forbeck Foundation (U.S.), the Neuroblastoma
Society (U.K.), the Norwegian Cancer Society, the
Gibraltar Neuroblastoma Society, Keio University
Medical Society, Tokyo (Japan), and the Ronald
McDonald Children’s Charity Fund (U.S.).
Address for reprints: Hiroyuki Shimada, M.D.,
Ph.D., Department of Pathology and Laboratory
Medicine, Childrens Hospital Los Angeles, MS #43,
4650 Sunset Boulevard, Los Angeles, CA 90027.
Received July 24, 1998; revision received January
8, 1999; accepted February 25, 1999.
© 1999 American Cancer Society
BACKGROUND. The International Neuroblastoma Pathology Committee, which is
comprised of six member pathologists, was convened with the objective of pro-
posing a prognostically significant and biologically relevant classification based on
morphologic features of neuroblastic tumors (NTs) (i.e., neuroblastoma, ganglion-
euroblastoma, and ganglioneuroma).
METHODS. A total of 227 cases were reviewed. Consensus diagnoses from morpho-
logic features (criteria described separately) based on five of six or six of six
agreements by the reviewer pathologists were used for prognostic analysis. Prog-
nostic effects of morphology, both individual and in combination, taken in con-
junction with age (Shimada classification, histologic grade, and risk group), were
analyzed.
RESULTS. Approximately 99% of cases (224 of 227) had consensus diagnoses for
categorization: neuroblastoma (Schwannian stroma-poor), 190 cases; ganglioneu-
roblastoma, intermixed (Schwannian stroma-rich), 5 cases; ganglioneuroma
(Schwannian stroma-dominant) maturing, 1 case; ganglioneuroblastoma, nodular
(composite Schwannian stroma-rich/stroma-dominant and stroma-poor), 19 cas-
es; and NT-unclassifiable, 9 cases. For the NTs, subtype (93% consensus: undif-
ferentiated, 6 cases; poorly differentiated, 155 cases; and differentiated, 15 cases),
mitosis-karyorrhexis index (90% consensus: low, 94 cases; intermediate, 40 cases;
and high, 37 cases), mitotic rate (75% consensus: low, 89 cases; high, 50 cases; and
not determined, 4 cases), and calcification (100% consensus: yes, 110 cases and no,
80 cases) were recorded. Statistical analysis demonstrated that the Shimada clas-
sification system (90% consensus; 3-year event free survival: 85% for the group with
favorable histology and 41% for the group with unfavorable histology; P 5 0.31 3
1029) had a significantly stronger prognostic effect than individual features and
other combinations.
CONCLUSIONS. The International Neuroblastoma Pathology Classification, a sys-
tem based on a framework of the Shimada classification with minor modifications,
is proposed for international use in assessing NTs. Cancer 1999;86:364 –72.
© 1999 American Cancer Society.
KEYWORDS: neuroblastic tumors, histopathology, classification, prognosis, Shi-
mada system.
A cooperative international effort has been in progress, assessing
clinical and biologic characteristics of neuroblastic tumors (NTs)
(a group of tumors including neuroblastoma, ganglioneuroblastoma,
and ganglioneuroma) for the development of treatment strategies
based on a complete set of International Neuroblastoma Risk Groups.
The international collaboration to date has produced an International
Neuroblastoma Staging System and a set of International Neuroblas-
toma Response Criteria.1 The International Neuroblastoma Pathology
Committee (INPC), established in 1994, has been actively participat-
ing in this effort by proposing a pathology classification for interna-
tional use.
NTs once were described as “enigmatic” by many oncologists and

investigators because of their “unexpected” clinical
behaviors, such as involution/spontaneous regres-
sion, maturation, and aggressive progression. Recent
advances in clinical as well as basic research have
been successfully accumulating data that are allowing
us to describe NTs in a new way (i.e., NTs are “heter-
ogeneous” with their individual biologic properties
related closely to their unique clinical behavior2).
Historically, the principal morphologic feature
recognized to be of prognostic importance for NTs is
the degree of neuroblastic maturation toward gan-
glion cells.3,4 A number of histopathologic grading
systems of NTs have been proposed in the past de-
cades, yet to our knowledge not one has gained uni-
versal favor and application.5,6 Shimada et al.7 took a
new approach with their age-linked classification,
which divided NTs into Schwannian stroma-rich and
stroma-poor tumors. They also have introduced the
term “mitosis-karyorrhexis index” (MKI) for describ-
ing one of the prognostic indicators. Some modifica-
tions in this classification were published by Joshi et
al.,8,9 who suggested a high mitotic rate to be an un-
favorable prognostic factor and tumor-associated cal-
cification to be a favorable prognostic factor. It has
become apparent over the past decade that there are
important biologic attributes of neuroblastic tumors
that have impacted on our understanding of these
neoplasms. The challenge of any pathology classifica-
tion, in common with others in the past, is to formu-
late a reproducible and biologically meaningful sys-
tem.
This report, based on the first international col-
laboration on NTs, summarizes 4-year activities of the
INPC: i.e., 1) making consensus diagnoses according
to uniform criteria of morphologic features and 2)
testing the prognostic significance of the morphologic
features and their combination by using the consen-
sus diagnoses. The goals and objectives of the INPC
are to propose an International Neuroblastoma Pa-
thology Classification, which should be prognostically
significant, biologically relevant, and at the same time
highly reproducible and user friendly. Detailed criteria
of morphologic features on the NTs used in this study
have been summarized in a previous article.10 In the
current study we also present a recommendation/
guideline for surgical pathologists to use in their de-
scription and prognostic evaluation of the NTs.
MATERIALS AND METHODS
Materials
The patients used in the analysis were a subset of
patients registered by January 8, 1995 in two Chil-
dren’s Cancer Group studies: CCG-3881 (a study for
low and intermediate risk neuroblastoma) and CCG-
Neuroblastoma Pathology Classification/Shimada et al. 365
3891 (a study for high risk neuroblastoma). Therefore
this combined group of patients incorporates all risk
groups. At the time that sample was created, a total of
535 patients from the two studies were identified as
having pathology material available for review. A ran-
dom sample of 152 patients (28% of these 535 pa-
tients) was taken, and, in addition, 75 subjects who
had developed disease recurrence or died by January
1995 were added to the sample for a total of 227
patients in the analysis (150 patients from CCG-3881
and 77 patients from CCG-3891). This sampling
scheme commonly is used in epidemiologic studies
and is known as the “case– cohort” design.11 It was
used here to maximize the amount of information
available in the statistical analysis, while keeping the
total number of subjects limited to a number that
could be reviewed over a limited period of time (all
subjects were reviewed in 5 working days). Pathology
slides of these 227 cases were collected from the CCG
Neuroblastoma Pathology Repository, Childrens Hos-
pital Los Angeles, Los Angeles, California, and ran-
domly numbered for the review. The individual cases
had varying numbers of sections (range, 1–35 sections,
average 6 sections per case). The surgical pathology
reports from the contributing institutions were care-
fully reviewed centrally by the pathologist of record
(H.S.) to determine the eligibility of the cases for the
study. Among these cases, 16 patients received inten-
sive treatment with autologous bone marrow trans-
plantation (ABMT) prior to the time of analysis of this
study. The neuroblastoma study group for CCG cur-
rently is in the process of analyzing the results. There
was a significant difference in event free survival (EFS)
between ABMT and chemotherapy of slightly ,15%
for the highest risk group. This level of difference in so
few patients treated with ABMT did not affect our
results to any significant degree in this study.
Among these cases, 108 patients were age ,1 year
and 119 were age $1 year at the time of diagnosis.
There were 30 Stage 1 cases, 59 Stage 2 cases, 31 Stage
3 cases, 87 Stage 4 cases, and 20 Stage 4-S cases
included in this study. Forty-two tumors had ampli-
fied MYCN, 152 had nonamplified MYCN, and the
MYCN test was not performed in 33 cases.
Pathology Review
Morphologic features of NTs to be evaluated by the
review were determined after the first meeting in Los
Angeles, California in 1994. They were defined precisely
and described in a previous article.10 The pathology re-
view was a step-wise evaluation of morphologic features
without knowledge of clinical information. First, tumors
were classified into four different categories: neuroblas-
toma (Schwannian stroma-poor); ganglioneuroblas-
366 CANCER July 15, 1999 / Volume 86 / Number 2
toma, intermixed (Schwannian stroma-rich); ganglion-
euroma (Schwannian stroma-dominant); and ganglio-
neuroblastoma, nodular (composite Schwannian stro-
ma-rich/stroma-dominant and stroma-poor). Second,
for neuroblastoma and the Schwannian stroma-poor
area of ganglioneuroblastoma, nodular, the grade of
neuroblastic differentiation (undifferentiated, poorly dif-
ferentiated, and differentiating) was determined along
with the other specific features (3 classes for MKI [low,
intermediate, and high], two classes for mitotic rate [MR]
[low and high], and the presence of calcification [yes or
no]). Third, for ganglioneuroma, two subtypes (maturing
and mature) were distinguished.
The first review session for the pathology slides
from the 227 cases was performed by the 6 patholo-
gists (H.S., I.M.A., L.P.D., J.H., V.V.J., and B.R.) individ-
ually in Oslo, Norway in 1995. After clarifying details
regarding the morphologic features, the second review
session was held in Tokyo, Japan in 1996. The second
session was a group review, and the 5 pathologists
present (H.S., I.M.A., J.H., V.V.J., and B.R.) examined
the same case at the same time by using a multihead
microscope. After review and discussion of the indi-
vidual cases by the group review, each pathologist cast
a vote for his or her final evaluation regarding type,
subtype, grade of differentiation, MKI, MR, and calci-
fication of the given tumor tissue. One pathologist
(L.P.D.) was unable to participate in the group review.
Prognostic Analysis
A total of six diagnoses (including five diagnoses made
by the group review and one diagnosis made by indi-
vidual review [L.P.D.]) for each morphologic feature
were collected. Consensus diagnoses for the individ-
ual morphologic features were decided by five of six or
six of six agreements by the reviewer pathologists, and
used for the analysis of the prognostic significance.
Prognostic effects of the individual morphologic
features and prognostic groupings according to vari-
ous classification systems were analyzed. The classifi-
cation systems analyzed in this study were: 1) the
Shimada classification7 (the favorable histology group,
including: (a) neuroblastoma with low or intermediate
MKI diagnosed at age ,1.5 years, (b) neuroblastoma,
differentiating subtype with low MKI diagnosed be-
tween ages 1.5–5 years, (c) ganglioneuroblastoma, in-
termixed in any age, and (d) ganglioneuroma in any
age; and the unfavorable histology group, including:
(a) neuroblastoma with high MKI in any age; (b) neu-
roblastoma, undifferentiated and poorly differentiated
subtype diagnosed between ages 1.5–5 years; (c) neu-
roblastoma with intermediate MKI diagnosed between
ages 1.5–5 years; (d) neuroblastoma diagnosed at age
.5 years; and (e) ganglioneuroblastoma, nodular di-
agnosed at any age); 2) histologic grade8 (Grade 1:
tumors with “MR 5 low and calcification 5 yes”;
Grade 2: tumors with “MR 5 low and calcification 5
no” and “MR 5 high and calcification 5 yes”, and
Grade 3: tumors with “MR 5 high and calcification 5
no”); 3) risk group8 (low risk: all Grade 1 tumors and
Grade 2 tumors diagnosed at age . 12 months and
high risk: all Grade 3 tumors and Grade 2 tumors
diagnosed at age $ 12 months); 4) modified histologic
grade12 (modified Grade 1: tumors with “MKI 5 low or
intermediate and calcification 5 yes,” modified Grade
2: tumors with “MKI 5 low or intermediate and calci-
fication 5 no” and “MKI 5 high and calcification 5
yes,” modified Grade 3: tumors with “MKI 5 high and
calcification 5 no”); and 5) modified risk group12
(modified low risk: all modified Grade 1 tumors and
modified Grade 2 tumors diagnosed at age , 12
months and modified high risk: all modified Grade 3
tumors and modified grade 2 tumors diagnosed at
age $ 12 months). Prognostic evaluations by the his-
tologic grade and the risk group were applicable only
to the neuroblastoma (Schwannian stroma-poor) tu-
mors.
Prognostic effects of the individual features and
the prognostic groups according to the classification
systems described earlier were analyzed. EFS was used
as the endpoint in the analysis. Kaplan–Meier analysis
(appropriately modified for the case– cohort selection)
was used to produce estimates of the probability of
EFS as specific times from diagnosis. Multivariate
analysis of the joint significance of individual factors
and prognostic groups was performed by the modified
Cox regression technique.11 All P values reported were
the result of two-tailed tests. Chemotherapy differed
between the two studies and by risk groups defined by
stage and biology. Low risk patients (Stages 1, 2, and
4-S) generally received treatment with surgery alone.
Intermediate risk patients (biologically favorable Stage
3 and Stage 4 infants) received a course of moderate
therapy on CCG-3881 and high risk patients (biologi-
cally unfavorable Stage 3 or Stage 4 infants or Stage 4
patients age . 1 year) received intensive chemother-
apy and possibly ABMT.
RESULTS
Among the 227 cases, consensus diagnoses for categori-
zation were reached for 224 tumors (99%). There were
190 neuroblastoma (Schwannian stroma-poor) tumors;
5 ganglioneuroblastoma, intermixed (Schwannian stro-
ma-rich) tumors; 1 ganglioneuroma (Schwannian stro-
ma-dominant), maturing tumor; 19 ganglioneuroblas-
toma, nodular (composite Schwannian stroma-rich/
stroma-dominant and stroma-poor) tumors; and 9 NT,
unclassifiable cases (Table 1). Morphologic features of
 Neuroblastoma Pathology Classification/Shimada et al. 367

TABLE 1 TABLE 3
Categorization of 227 Neuroblastic Tumors by 6 Reviewer Prognostic Effects of the Morphologic Features in Neuroblastoma
Pathologists (Schwannian Stroma-Poor) Tumors

Categories No. of cases Morphologic features Survivala P value

Consensus diagnosis (99%)a Subtypes

Neuroblastoma (Schwannian stroma-poor) 190 Undifferentiated (a) 30% (a) vs. (b), P 5 0.11 3 1024
Ganglioneuroblastoma, intermixed (Schwannian stroma-rich) 5 Poorly differentiated (b) 73% (a) vs. (c), P 5 0.96 3 1024
Ganglioneuroma (Schwannian stroma-dominant) 1 Differentiating (c) 77% (b) vs. (c), P 5 0.81
Ganglioneuroblastoma, nodular (composite Schwannian MKI
stroma-rich/stroma-dominant and stroma-poor) 19 Low (a) 82% (a) vs. (b), P 5 0.0069
NT, unclassifiableb 9 Intermediate (b) 65% (a) vs. (c), P 5 0.77 3 1029
Disagreement (1%) 3 High (c) 23% (b) vs. (c), P 5 0.026
Mitotic rate
NT: neuroblastic tumor. Low 83% P 5 0.0019
a
Consensus diagnosis: based on five of six or six of six agreements by the reviewer pathologists. High 58%
b
Criteria presented previously.10 Calcification
Yes 76% P 5 0.019
No 62%
TABLE 2
Morphologic Features of 190 Cases of Neuroblastoma (Schwannian MKI: mitosis-karyorrhexis index.
Stroma-Poor) by 6 Reviewer Pathologists a
Survival is shown as expected 3-year progression free survival.

Neuroblastoma (Schwannian stroma-poor) No. of cases

of 190) had consensus diagnoses for calcification (yes,
Subtype
Consensus diagnosis (93%)a 110 cases and no, 80 cases). With regard to the morpho-
Undifferentiated 6 logic features of the stroma-poor portion of 19 ganglion-
Poorly differentiated 155 euroblastoma, nodular tumors, 95% (18 of 19) had con-
Differentiating 15 sensus diagnoses for subtyping (12 poorly differentiated
Disagreement (7%) 14
and 5 differentiated tumors with 1 not evaluable case);
MKI
Consensus diagnosis (90%)a 95% (18 of 19) had consensus diagnoses for MKI (14 low
Low (, 100/5000) 94 MKI tumors, 2 intermediate MKI tumors, and 2 high MKI
Intermediate (100–200/5000) 40 tumors); 89% (17 of 19) had consensus diagnoses for MR
High (. 200/5000) 37 (14 low MR tumors and 1 high MR tumor with 2 not
Disagreement (10%) 19
evaluable cases); and 100% (19 of 19) had consensus
Mitotic rate
Consensus diagnosis (75%)a diagnoses for calcification (yes, 15; no, 3; and one not
Low (# 10/10 HPF) 89 evaluable case).
High (. 10/10 HPF) 50 Among these patients, 117 had events during their
Not determined 4 clinical course (follow-up period range, 0.68 –57.87
Disagreement (25%) 47
months; median, 25.11 months); 42 showed disease
Calcification
Consensus diagnosis (100%)a progression but were alive at last follow-up, 66 showed
Yes 110 disease progression and died of tumor, and 9 died of
No 80 tumor soon after diagnosis. Prognostic significance of
Disagreement (0%) 0 the individual morphologic features for the neuroblas-
toma tumors based on the consensus diagnoses is
MKI: mitosis-karyorrhexis index; HPF: high-power fields.
a
Consensus diagnosis: based on five of six or six of six agreements by the reviewer pathologists. listed in Table 3. Nine cases of NT, unclassifiable were
excluded from the prognostic analysis. The individual
features had prognostic effects in EFS rates, with var-
the 190 neuroblastoma tumors are summarized in Table ious degrees of significance. It was noted that those
2: 93% (176 of 190) had consensus diagnoses for their tumors of undifferentiated subtype had a significantly
subtyping based on grade of neuroblastic differentiation lower 3-year EFS of 30%. In addition, those tumors
(6 undifferentiated, 155 poorly differentiated, and 15 dif- with high MKI showed a significantly lower 3-year EFS
ferentiating tumors). 90% (171 of 190) had consensus of only 23%. MR had the lowest consensus rate and
diagnoses for MKI (94 low MKI tumors, 40 intermediate did not add any significant prognostic information to
MKI tumors, and 37 high MKI tumors). Approximately that of MKI.
75% (143 of 190) had consensus diagnoses for MR (89 Because the number of cases of ganglioneuroblas-
low MR tumors and 50 high MR tumors), and 100% (190 toma, nodular was limited, analysis of prognostic ef-
368 CANCER July 15, 1999 / Volume 86 / Number 2

TABLE 4
Prognostic Effects by Grouping

Prognostic grouping Survivala P value

Shimada classification (consensusb

90%)
Favorable histology (N 5 103) 85% P 5 0.31 3 1029
Unfavorable histology (N 5 93) 41%
Histologic grade (consensusb 73%)
Grade 1 (a) (N 5 62) 84% (a) vs. (b), P 5 0.12
Grade 2 (b) (N 5 57) 73% (a) vs. (c), P 5 0.0088
Grade 3 (c) (N 5 20) 58% (b) vs. (c), P 5 0.18
Risk group (consensusb 73%)
Low risk (N 5 85) 85% P 5 0.41 3 1023
High risk (N 5 54) 59%
Histologic grade, modified (consensusb
90%) FIGURE 1. Kaplan–Meier curves (event free survival rates) for tumors with
Grade 1 (a) (N 5 85) 78% (a) vs. (b), P 5 0.36
favorable and unfavorable histology according to the Shimada System (P 5
Grade 2 (b) (N 5 63) 70% (a) vs. (c), P 5 0.39 3 1026
Grade 3 (c) (N 5 23) 18% (b) vs. (c), P 5 0.13 3 1024 0.31 3 1029). Three-year 95% confidence intervals for favorable histology
Risk group, modified (consensusb 90%) tumors were 0.7271, 0.9771 and those for tumors with unfavorable histology
Low risk (N 5 116) 80% P 5 0.53 3 1025 were 0.1901, 0.6313. *#pts. at risk (FH/UH): numbers of patients at risk at
High risk (N 5 55) 42% different time points, FH: favorable histology; UH: unfavorable histology.

The Shimada Classification was applied to all evaluable neuroblastic tumors (N 5 218). Histologic
grade; risk group; histologic grade, modified; and risk group, modified were applied only to the
neuroblastoma (Schwannian stroma-poor) tumors (N 5 190).
a
Survival is shown as the expected 3-year event free survival.
b
Consensus was based on five of six or six of six agreements by the reviewer pathologists.
fects for the individual morphologic features of their
Schwannian stroma-poor portion was not performed.
Two patients with ganglioneuroblastoma, intermixed
had progression of disease after incomplete resection,
but all 5 patients in this category and 1 patient with
ganglioneuroma, maturing were alive after follow-up
periods ranging from 1 year and 8.5 months to 4 years
and 8 months.
Impacts on EFS rates by the different prognostic
classification systems (i.e., the Shimada classification
[consensus diagnoses for 90% cases], histologic grade
[consensus diagnoses for 73% cases], risk group [con-
sensus diagnoses for 73% cases], modified histologic
grade [consensus diagnoses for 90% cases], and mod-
ified risk group [consensus diagnoses for 90% cases])
are summarized in Table 4. All the classifications had
prognostic effects of various significance. Among
them, the Shimada classification (age, categorization
by Schwannian stromal development, subtyping by
grade of neuroblastic differentiation, and MKI) and
modified risk group (age, calcification, and MKI), had
a very strong impact on the EFS rates of the patients
with NTs, and showed very high consensus rates. Mul-
tivariate analysis disclosed that 1) the Shimada classi-
fication (Fig. 1) (103 tumors in the favorable histology
group and 93 tumors in the unfavorable histology
group) was the strongest and could add more prog-
nostic information to the prognostic effect of the mod-
ified risk group (P 5 0.0012); 2) “calcification,” one of
the prognostic indicators in the modified risk group,
did not add any significant prognostic information to
the Shimada classification; and 3) the prognostic ef-
fect of the modified risk group was influenced signif-
icantly by MKI alone.
Additional analyses were performed for the signif-
icance of age factor on the Shimada classification sys-
tem. First, the system without age factor gave signifi-
cantly lower prognostic effects than the original
classification (P 5 0.54 3 1023). Then, Kaplan–Meier
curves generated by different, month-by-month, age
cutoffs from 12 months to 24 months demonstrated
that the cutoff at 18 months of age used in the original
classification gave the most significant information for
prognostic evaluation. For example, P values indicat-
ing better prognostic effect of the cutoff at 18 months
by direct comparison with other months were: 0.63 3
1029 for 12 months versus 18 months, 0.052 for 14
months versus 18 months, 0.072 for 16 months versus
18 months, 0.076 for 20 months versus 18 months,
0.019 for 22 months versus 18 months, and 0.11 for 24
months versus 18 months.
Because undifferentiated subtype of neuroblas-
toma had a very low EFS rate, all six cases were listed
in Table 5. It was noted that all cases were classified
into an unfavorable histology group according to the
Shimada classification system.
Table 6 shows distribution of age, clinical stage,
and MYCN status by the International Neuroblastoma
Pathology Classification (Shimada system; a total of
 Neuroblastoma Pathology Classification/Shimada et al. 369

TABLE 5 ter prognostic information than the modified risk

List of 6 Cases with Undifferentiated Subtype of Neuroblastoma group12 whose prognostic effect largely is influenced
(Schwannian Stroma-Poor)
by MKI alone. The Committee members were able to
Shimada produce consensus diagnoses in 90% of cases in the
Age (mos) MKI Mitotic rate Calcification classification classification with very high levels of concordance of
the individual morphologic features. Although the
17 High NCD No Unfavorable current study was not designed to test reproducibility,
15 High NCD No Unfavorable
description of the criteria reported in a preceding ar-
4 High NCD No Unfavorable
24 Intermediate Low No Unfavorable ticle10 would assure pathologists, as it did the Com-
10 High High Yes Unfavorable mittee members, a high consensus/concordance in
48 Low Low No Unfavorable making histologic evaluation of NTs. The Committee
currently is planning to make an atlas of NT histopa-
MKI: mitosis-karyorrhexis index; NCD: no consensus diagnosis.
thology for international use. The International Neu-
roblastoma Pathology Classification is to serve a stan-
TABLE 6 dardized formula of prognostic assessment by
Prognostic Groups According to the International Neuroblastoma pathology evaluation of the NTs. The classification
Pathology Classification: Case Distribution by Nonmorphologic
also can be used as a tool for comparison of histologic
Prognostic Factors
features with genetic and molecular-biologic proper-
International Neuroblastoma Pathology ties of the tumors.
Classification (Shimada System) Biologic relevance of the morphologic features
used in the International Neuroblastoma Pathology
Favorable histology Unfavorable histology P value
Classification, such as Schwannian stromal develop-
Age ment in the age-linked maturational sequence of the
, 1 year 72 22 NTs, nodular formation in the ganglioneuroblastoma,
$ 1 year 31 71 , 0.0001 nodular (composite Schwannian stroma-rich/stroma-
Clinical stage dominant and stroma-poor) tumors, and grade of
1 1 2 1 4-S 73 20
neuroblastic differentiation and MKI in the neuroblas-
314 30 73 , 0.0001
MYCNa toma (Schwannian stroma-poor) tumors, has been
Nonamplified 84 48 discussed in a previous article.10 It should be noted
Amplified 6 32 , 0.0001 again here that there is a reproducible correlation
a
between a high MKI and adverse clinical and biologic
MYCN status was tested in 170 of 196 available cases.
(MYCN amplification) manifestations.13 It is specu-
lated that the karyorrhectic process in tumor cells with
amplified MYCN could be different biologically from
196 cases available for the analysis). It was apparent the apoptotic process in those with potential for dis-
that the tumors in the favorable histology group were ease regression.10 Consensus rate and prognostic im-
associated significantly with biologically favorable fac- pact of the MR were not sufficient to be included in
tors such as younger age at diagnosis (, 1 year), the prognostication. Calcification, in spite of its 100%
nonadvanced clinical stage (Stage 1 or 2) or Stage 4-S, consensus, did not appear to have an isolated prog-
and nonamplified MYCN, whereas the tumors in the nostic impact.
unfavorable histology group frequently presented with The results of statistical analyses clearly demon-
biologically unfavorable factors such as older age at strated that age is critical in the prognostic evaluation
diagnosis ($ 1 year), advanced clinical stage (Stage 3 of NT morphology. According to the Shimada system,
or 4), and amplified MYCN. there are two cutoffs for the age factor in neuroblas-
toma tumors: the prognostic significance of the cutoff
DISCUSSION age of 1.5 years was shown in this study, and that of
Based on the detailed definition of morphologic fea- another age cutoff of 5 years was confirmed by a
tures10 and the statistical analysis of pathology review recent study using a large series of cases.14 It becomes
data, the International Neuroblastoma Pathology evident that two morphologic indicators (i.e., grade of
Classification is proposed by adopting the Shimada neuroblastic differentiation and MKI) in neuroblas-
classification system7 with minor modifications (Table toma tumors have different prognostic effects de-
7). This is an age-linked classification system that pending on the patient’s age at diagnosis. The data
prognostically is the most significant and biologically also support hypotheses used in the Shimada system:
relevant. The Shimada system offers significantly bet- 1) neuroblastic cells with a maturational potential re-
370 CANCER July 15, 1999 / Volume 86 / Number 2

TABLE 7
Prognostic Evaluation of Neuroblastic Tumors According to the International Neuroblastoma Pathology Classification (Shimada System)

International Neuroblastoma Pathology classification Original Shimada classification Prognostic group

Neuroblastoma (Schwannian stroma-poor)a Stroma-poor

Favorable Favorable Favorable
, 1.5 yrs Poorly differentiated or differentiating
& low or intermediate MKI tumor
1.5–5 yrs Differentiating & low MKI tumor
Unfavorable Unfavorable Unfavorable
, 1.5 yrs a) undifferentiated tumorb
b) high MKI tumor
1.5–5 yrs a) undifferentiated or poorly
differentiated tumor
b) intermediate or high MKI tumor
$5 yrs All tumors
Ganglioneuroblastoma, intermixed (Schwannian stroma-rich) Stroma-rich Intermixed Favorablec
(favorable)
Ganglioneuroma (Schwannian stroma-dominant)
Maturing Well differentiated (favorable) Favorablec
Mature Ganglioneuroma
Ganglioneuroblastoma, nodular (composite Schwannian stroma-rich/ Stroma-rich nodular (unfavorable) Unfavorablec
stroma-dominate and stroma-
poor)

MKI: mitosis-karyorrhexis index.

a
Subtypes of neuroblastoma were described in detail elsewhere.10
b
Rare subtype, especially diagnosed in this age group. Further investigation and analysis required.
c
Prognostic grouping for these tumor categories is not related to patient age.

quire an in vivo latent period for their demonstration
of histologic evidence of differentiation and 2) there is
a certain allowance for mitotic and karyorrhectic ac-
tivities of neuroblastic cells in infants and younger
children. Recent data suggest a biologically significant
correlation between higher expression of nerve growth
factor receptors and an age-linked morphologic differ-
entiation of the neuroblastic cells in NTs.15
As illustrated in Table 7, the International Neuro-
blastoma Pathology Classification distinguishes favor-
able and unfavorable histology groups according to
the Shimada system.7 A conceptual framework of age-
linked maturation is used for identifying tumors in the
former group, although tumors in the latter group
have morphologic indicators suggesting their aggres-
sive growth. Tumors in the favorable histology group
are within an age-linked maturational sequence from
poorly differentiated (age , 1.5 years) to differentiat-
ing (age , 5 years) neuroblastoma to ganglioneuro-
blastoma, intermixed and to ganglioneuroma. The
neuroblastoma tumors in this group should have low
(age , 5 years) or up to intermediate (age , 1.5 years)
MKI. By contrast, tumors in the unfavorable histology
group have immature histologies for the age of the
patients and are classified into an undifferentiated (in
any age) or poorly differentiated subtype (age $ 1.5
years) or any subtype (age $ 5 years) of the neuroblas-
toma. Among the neuroblastoma tumors, those with
high MKI (in any age) or intermediate (age $ 1.5 years)
also qualify as having an unfavorable histology. Neu-
roblastic nodular formation in the ganglioneuroblas-
toma, nodular tumor also is considered a sign of ag-
gressive growth.
As shown in Table 6, tumors in the favorable his-
tology group were associated with biologically favor-
able nonmorphologic prognostic factors, whereas tu-
mors in the unfavorable histology group were related
closely with biologically unfavorable, nonmorphologic
prognostic factors. However, histopathologic evalua-
tion according to the Shimada system has been shown
to provide additional prognostic information beyond
the factors of age, stage, and MYCN status.13,16
The only differences between the original Shi-
mada classification7 and the International Neuroblas-
toma Pathology Classification are 1) to subdivide the
“undifferentiated” subtype in the former classification
into two subtypes of undifferentiated and poorly dif-
ferentiated in the latter classification, and 2) to change
the name of “stroma-rich, well differentiated” in the
former classification to “ganglioneuroma, maturing”
in the latter classification. The undifferentiated sub-
type is created partly due to a practical reason10 and
partly due to a possible prognostic implication: it con-
tains a small number of cases in our series whose

prognosis was extremely poor. A similar poor progno-
sis has been reported in a previous series from the
Pediatric Oncology Group (POG).9 All undifferentiated
neuroblastoma tumors in our series, as well as a series
from the POG, were classified into the unfavorable
histology group according to the Shimada system.
High MKI, another powerful indicator of poor prog-
nosis among individual morphologic features, is alone
a prognostic indicator for predicting aggressive tumor
behavior in the Shimada system. Lastly, ganglioneu-
roma in the International Neuroblastoma Pathology
Classification is accompanied by a term of Schwan-
nian stroma-dominant in parenthesis.
Currently intergroup studies by CCG and POG in
the North America are using the original Shimada
classification for patient stratification and protocol
assignment. We believe the International Neuroblas-
toma Pathology Classification (Shimada system) will
become one of the official prognostic indicators for
analyzing a large number of cases in their studies after
the publication. The prognostic significance of the
classification also will be tested in European countries
as well as in Japan.
Guidelines for Description in the Surgical
Pathology Report
Following are recommendations for the surgical pa-
thologist in preparing pathology reports on NT.
The morphologic description of the NT should
include information regarding the amount of Schwan-
nian stroma and the presence or absence of macro-
scopic nodularity, indicating a diagnosis of composite
tumor. The degree of neuroblastic differentiation and
MKI are documented for the neuroblastoma and for
the stroma-poor portion of the ganglioneuroblastoma,
nodular. The description also includes a note regard-
ing the presence or absence of calcification for future
analysis (calcification is the only morphologic indica-
tor used in the modified risk group12 but is not in-
cluded in the Shimada system7), but not for the cur-
rent prognostic evaluation. It also should indicate the
percentage of neuroblastic/ganglion cells versus
Schwannian and other normal cells in the sample
tissue(s) used for genetic/biologic studies (see recom-
mendations for tissue handling and tissue/cell preser-
vation in the article by Shimada et al.10).
The pathologic diagnosis includes a morphologic
description only, expressing the category and subtype
of the NT. If tumor cells were observed beyond the
surgical margin(s) of resection, it should be indicated
in the diagnosis without a conclusion as to whether
the residual tumor is microscopic or macroscopic.
In comments accompanying the diagnosis,
prognostic evaluation of morphologic features is de-
Neuroblastoma Pathology Classification/Shimada et al. 371
scribed in conjunction with the patient’s age. In this
section, the morphologic diagnosis also needs to be
discussed in relation to known data of biologic im-
portance. Results of the genetic and molecular biol-
ogy studies (i.e., MYCN status, chromosome 1p de-
letions, DNA content, etc.) can be included in an
integrated prognostic evaluation. MYCN amplifica-
tion and deletions at chromosome 1p both suggest
an unfavorable prognosis.17–19
With regard to the prognostic evaluation based on
a limited amount of tissue from a pathology specimen,
careful discussion should be required with oncology
team members including an oncologist, radiologist,
and surgeon (see problematic cases for histologic eval-
uation in the article by Shimada et al.10). This is very
important, especially when differential diagnoses in-
clude ganglioneuroblastoma, intermixed (Schwannian
stroma-rich), ganglioneuroma (Schwannian stroma-
dominant), and ganglioneuroblastoma, nodular (com-
posite Schwannian stroma-rich/stroma-dominant
and stroma-poor). Prognostic evaluation based on the
review of a limited amount of specimen should be
offered only when all the team members are con-
vinced that the submitted specimen is representative
of the characteristics of the entire tumor tissue. Even
then, the report should indicate clearly that the prog-
nostic evaluation was made based on a limited
amount of tissue sample. With regard to neuroblas-
toma, tumor tissue containing at least 5000 viable
neuroblastic cells from multiple microscopic fields for
the assessment of MKI and grade of neuroblastic dif-
ferentiation is required for prognostic evaluation.
REFERENCES
1. Brodeur GM, Pritchard J, Berthold F, Carlsen NLT, Castel V,
Castleberry RP, et al. Revisions of the International Criteria
for Neuroblastoma Diagnosis, Staging, and Response to
Treatment. J Clin Oncol 1993;11:1466 –77.
2. Brodeur GM, Castleberry RP. Neuroblastoma. In: Pizzo PA,
Poplack DG, editors. Principles and practice of pediatric
oncology. 3rd edition. Philadelphia: Lippincott-Raven Pub-
lishers, 1997:761–97.
3. Beckwith JB, Martin RF. Observations on the histopathology
of neuroblastomas. J Pediatr Surg 1968;3:106 –10.
4. Hughes M, Marsden HB, Palmer MK. Histologic patterns of
neuroblastoma related to prognosis and clinical staging.
Cancer 1974;34:1706 –11.
5. Dehner LP. Classic neuroblastoma: histopathologic grading
as a prognostic indicator, The Shimada System and its pro-
genitors. Am J Pediatr Hematol Oncol 1988;10:143–54.
6. Joshi VV, Tsongalis GJ. Correlation between morphologic
and nonmorphologic prognostic markers of neuroblastoma.
Ann N Y Acad Sci 1997;824:71– 83.
7. Shimada H, Chatten J, Newton WA, Sachs N, Hamoudi AB,
Chiba T, et al. Histopathologic prognostic factors in neuro-
blastic tumors: definition of subtypes of ganglioneuroblas-
toma and an age-linked classification of neuroblastomas.
J Natl Cancer Inst 1984;73:405–16.
372 CANCER July 15, 1999 / Volume 86 / Number 2
8. Joshi VV, Cantor AB, Altshuler A, Larkin EW, Neill JSA, Shus-
ter JJ, et al. Age-linked prognostic categorization based on a
new histologic grading system of neuroblastomas: a clinical
pathologic study of 211 cases from the Pediatric Oncology
Group. Cancer 1992;69:2197–211.
9. Joshi VV, Cantor AB, Altshuler A, Larkin EW, Neill JSA, Shus-
ter JJ, et al. Recommendation for modification of terminol-
ogy of neuroblastic tumors and prognostic significance of
Shimada classification: a clinicopathologic study of 213
cases. Cancer 1992;69:2183–96.
10. Shimada H, Ambros I, Dehner LP, Hata J, Joshi VV, Roald B.
Terminology and morphologic criteria of neuroblastic tu-
mors: recommendations by the International Neuroblas-
toma Pathology Committee. Cancer 1999;86:348 – 62.
11. Prentice RL. A case-cohort design for epidemiologic cohort
studies and disease prevention trials. Biometrika 1986;73:1–11.
12. Joshi VV, Rao PV, Cantor AB, Altshuler G, Shuster JJ, Castle-
berry RP. Modified histologic grading of neuroblastoma by
replacement of mitotic rate with mitosis karyprrhexis index:
a clinicopathologic study of 223 cases from the Pediatric
Oncology Group. Cancer 1996;77:1582– 8.
13. Shimada H, Stram D, Chatten J, Joshi VV, Hachitanda Y,
Brodeur GM, et al. Identification of subsets of neuroblasto-
mas combined histopathologic and N-myc analysis. J Natl
Cancer Inst 1995;87:1470 – 6.
14. Monforte-Munoz H, Kawakami T, Stram D, Gerbing MA,
Matthay K, Lukens J, et al. Neuroblastoma in children over
5 years of age: recognition of a rare and aggressive subset
with “unconventional” morphology. Mod Pathol 1997;10(1):
4P.
15. Peters J, Miyake M, Seeger RC, Cai B, Yao D, Hong CM, et al.
Splicing pattern of the high affinity nerve growth factor
receptor (trkA) transcript in neuroblastomas: correlation
with histopathology and clinical outcome. Lab Invest 1995;
72(1):144A.
16. Matthay KK, Perez C, Seeger RC, Brodeur GM, Shimada H,
Atkinson JB, et al. Successful treatment for stage III neuro-
blastoma based on prospective biologic staging: a Childrens
Cancer Group Study. J Clin Oncol 1998;16:1256 – 64.
17. Brodeur GM. Molecular basis for heterogeneity in human
neuroblastoma. Eur J Cancer 1995;31A(4):505–10.
18. Ambros PF, Ambros IM, Strehl S, Bauer S, Luegmayr A,
Kovar H, et al. Regression and progression in neuroblas-
toma. Does genetics predict tumour behavior? Eur J Cancer
1995;31A(4):510 –5.
19. Caron H, Van Sluis P, de Kraker J, Bokkerink J, Egeler M,
Laureys G, et al. Allelic loss of chromosome 1p as a predictor
of unfavourable outcome in patients with neuroblastoma.
N Engl J Med 1996;334:225–30.